{"slug": "i-built-a-free-byok-ai-coding-agent-ide-that-runs-fully-offline-on-local-llms", "title": "I built a free, BYOK AI coding agent IDE that runs fully offline on local LLMs", "summary": "A developer built Forger, a free, source-available Electron IDE that bundles a file explorer, Monaco editor, AI chat, Git, and a TTY terminal into a single executable with a multi-step agent loop gated behind approval dialogs. The tool supports bring-your-own-key Gemini access as well as fully offline operation via Ollama, sending requests only to the user-configured LLM endpoint with no telemetry. The developer says it was built with AI in under three days and is released under FSL-1.1-MIT, converting to MIT two years after each release.", "body_md": "I built **Forger**, a standalone Electron app that bundles a file\n\nexplorer, Monaco editor, AI chat, Git, and a real TTY terminal into one\n\nwindow — with an agent loop that lists, reads, searches, and edits\n\nfiles across multiple steps, all behind approval dialogs.\n\nRepo: [https://github.com/cuculhart/forger-ide](https://github.com/cuculhart/forger-ide)\n\nSetup guide: [https://cuculhart.com/forger-howto.en.html](https://cuculhart.com/forger-howto.en.html)\n\nNot a VS Code fork, not an extension — a single exe.\n\n## \n  \n  \n  Why I built it\n\nDoing contract work, I kept hitting the same walls:\n\n- \n**Confidential code can't go to vendor clouds.** In contract work you\nhandle code and engineering knowledge that simply may not pass\nthrough someone else's servers.\n- \n**Orgs that can't approve $20/mo subscriptions.** There are still\nplenty of organizations where a monthly AI IDE subscription is a\nhard sell.\n- \n**VS Code extension agents get tangled.** Extensions are convenient,\nbut they also collide with each other — Copilot interference being\nthe usual suspect.\n- \n**Unapproved rewrites are a review burden.** When the AI silently\nrewrites code, the human reviewer pays the cognitive cost. So every\nwrite, diff edit, and command goes through an approval dialog, and\ncheckpoint/rollback undoes only the AI's changes.\n- \n**Docker-in-Docker tools are heavy.** Nice, but a hard setup for\nbeginners. I wanted a single exe that just runs.\n- \n**Beginners need visible panes.** Explorer, editor, console, and AI\nchat as separate, obvious panes is simply easier to learn.\n\nNothing matched all of this — so I built it, with AI, in under 3 days.\n\n## \n  \n  \n  Features\n\n- \n**Free + BYOK** — bring your own Gemini API key, pay Google directly\nfor what you use. No subscription, no markup.\n- \n**Fully offline option** — point it at Ollama (Gemma, Qwen, ...) and\nevery request goes to localhost only. Your code never leaves the\nmachine.\n- \n**No telemetry** — the only network calls go to the LLM endpoint you\nconfigure. Source-available, so you can verify that yourself.\n- \n**Human-in-the-loop** — writes are confined to the project root;\nfile creation, diff edits, and shell commands all require approval.\n- \n**LiteLLM proxy support** — real API keys never sit on the client.\n- \n**Token-efficient context** — the AI gets the file tree and fetches\ncontents on demand via`// GREP:` /`// READ_FILE:` commands.\n- \n**Cheap by default** — Gemini 3.8 Flash is at introductory pricing\n(half the standard rate) through the end of 2026, and even cheaper\nmodels like 3.5 Flash-Lite are selectable.\n\n## \n  \n  \n  Tech stack\n\nElectron / React / TypeScript / Monaco Editor / Gemini API & Ollama /\n\nVite\n\n## \n  \n  \n  License\n\nFSL-1.1-MIT (source-available). Free to use, modify, fork, and\n\nredistribute for any non-competing purpose — including commercial work\n\ninside your company. The one restriction: you can't sell it (or a\n\nrenamed clone) as a competing product. Each release automatically\n\nconverts to MIT two years after publication.\n\n## \n  \n  \n  Honest limitations\n\nIt's chat/agent-centric (no inline completions), there's no codebase\n\nindex (tree + on-demand reads instead), editor tabs don't exist yet,\n\nand it's Windows-first with the UI in English/Japanese.\n\n## \n  \n  \n  Roadmap\n\nConsidering PDF/image recognition, making the agent loop more reliable\n\non small local models, and chat improvements (multiple chat tabs).\n\nFeedback and bug reports welcome via\n\n[GitHub Issues](https://github.com/cuculhart/forger-ide/issues).", "url": "https://wpnews.pro/news/i-built-a-free-byok-ai-coding-agent-ide-that-runs-fully-offline-on-local-llms", "canonical_source": "https://dev.to/cuculhart/i-built-a-free-byok-ai-coding-agent-ide-that-runs-fully-offline-on-local-llms-4bdb", "published_at": "2026-09-23 08:33:21+00:00", "updated_at": "2026-09-23 08:58:37.560356+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools", "large-language-models", "ai-products"], "entities": ["Forger", "Electron", "Monaco Editor", "Gemini API", "Ollama", "LiteLLM", "GitHub", "Vite"], "alternates": {"html": "https://wpnews.pro/news/i-built-a-free-byok-ai-coding-agent-ide-that-runs-fully-offline-on-local-llms", "markdown": "https://wpnews.pro/news/i-built-a-free-byok-ai-coding-agent-ide-that-runs-fully-offline-on-local-llms.md", "text": "https://wpnews.pro/news/i-built-a-free-byok-ai-coding-agent-ide-that-runs-fully-offline-on-local-llms.txt", "jsonld": "https://wpnews.pro/news/i-built-a-free-byok-ai-coding-agent-ide-that-runs-fully-offline-on-local-llms.jsonld"}}